Research map / Video / Action recognition
Uncertainty-Boosted Robust Video Activity Anticipation
Zhaobo Qi, Shuhui Wang, Weigang Zhang, Qingming Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add uncertainty-boosted learning to RULSTM | RULSTM · EK100 validation (Act) | top-5 recall (%): 14.0 -> 16.6 | Improved |
| add uncertainty-boosted learning to DCR | DCR · EK100 validation (Act) | top-5 recall (%): 18.3 -> 19.1 | Improved |
| add uncertainty-boosted learning to ActionBanks | ActionBanks · EK55 validation | top-5 accuracy (%): 28.6 -> 30.2 | Improved |
| add distribution adjustment using uncertainty (ablation) | Ub-Baseline · EK55 validation (1.0s) | top-5 accuracy (%): 31.6 -> 32.1 | Improved |
| add precise target activity label construction (ablation) | Ub-Baseline · EK55 validation (1.0s) | top-5 accuracy (%): 31.2 -> 32.1 | Improved |
| add sample-wise relative uncertainty learning (ablation) | Ub-Baseline · EK55 validation (1.0s) | top-5 accuracy (%): 32.0 -> 32.1 | Improved |
| add temporal relative uncertainty learning (ablation) | Ub-Baseline · EK55 validation (1.0s) | top-5 accuracy (%): 31.6 -> 32.1 | Improved |
| add class-balanced loss to DCR | DCR · EK100 validation (Act) | top-5 recall (%): 14.6 -> 12.7 | Hurt |
| use uncertainty-based labels instead of class-balanced loss for tail classes | Ub-DCR · EK100 tail classes (Act) | top-5 recall (%): 11.4 -> 13.5 | Improved |
| add Gaussian noise to features (eta 0 to 10) | Ub-DCR · EK55 validation | top-5 accuracy (%): 34.2 -> 11.4 | Hurt |
| add Monte Carlo Dropout model uncertainty to Ub-DCR | Ub-DCR · EK55 validation | top-5 accuracy (%): 34.2 -> 34.7 | Improved |
| add Monte Carlo Dropout model uncertainty to Ub-DCR | Ub-DCR · EK100 validation (TSN) | top-5 accuracy (%): 14.8 -> 14.7 | No clear effect |
Closest papers by meaning
- Time-Conditioned Action Anticipation in One Shot (CVPR 2019 · similarity 0.85)
- Towards Understanding Future: Consistency Guided Probabilistic Modeling for Action Anticipation (AAAI 2024 · similarity 0.85)
- Rethinking Learning Approaches for Long-Term Action Anticipation (ECCV 2022 · similarity 0.85)
- Encouraging LSTMs to Anticipate Actions Very Early (ICCV 2017 · similarity 0.85)
- Hierarchical Multi-Source Uncertainty Aggregation for Interactive Video Captioning (AAAI 2025 · similarity 0.83)
- MANTA: Diffusion Mamba for Efficient and Effective Stochastic Long-Term Dense Action Anticipation (CVPR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
